Understanding and taxonomy of uncertainty in modeling, simulation, and risk profiling for border control automation

Abstract

This paper addresses the problem of trust in Modeling and Simulation (M&S) technologies, and uncertainty in applications to homeland security. The key goal of this paper is an extension of the notion of trusted M&S techniques for traveler risk assessment in mass-transit applications such as e-borders. Theories of uncertainty suggest that different understandings of uncertainty result in different mechanisms of its reduction. We show that a taxonomy of uncertainty that is accepted in philosophical studies, as well as the NATO methodology of uncertainty assessment (known as the Admiralty Code), can be useful in M&S. This paper overviews various approaches to M&S and focuses on a framework that is based on multi-source fusion mechanisms using Dempster–Shafer (DS) theory. The DS metric is useful for the development of simulators, recommender machines, and risk profilers when expert knowledge is given in an imprecise form. The difference between the Bayesian and DS metrics is introduced via a demonstrative experiment from the area of traveler risk assessment using a biometric-enabled watchlist.

Cognitive checkpoint: Emerging technologies for biometric-enabled watchlist screening

“…• According to resent studies [49], [50], [91] using the biometric-enabled watchlists (Level IV) is imperative. This may, however, decrease the performance.…”

“…In [39], the design goal is to minimize risks that "all the holes of a "Swiss Cheese" model do not line up". Some efforts to improve the performance of layered security have been reported with respect to: 1) topology of the waiting queuing lines [53] and checkpoint flow models [52]; 2) optimization of the passenger flows [64], [89]; 3) development of security measures [16], [78], including measures of the cost of travel time variability [24]; 4) cost-efficient minimization of security layers [77]; 5) modeling and simulation using a multi-state model of service [59], analytic hierarchy model [92] and hybrid models; for example, combining analytic hierarchy model and others, such as Dempster-Shafer [6], and Bayesian [58]; 6) traveler authentication and risk assessment, in particular, using multi-metric causal models [50], [91]; and cognitive agent models [45].…”

“…Misidentification remains a common occurrence due to shortcomings in current watchlist technologies and provides the principal motivation for this paper. Contemporary checkpoint is a cyber-physical computational platform that includes traveler authentication [23], [29], [48] using e-passport/ID [5], watchlist screening [9], [51], [86], concealed object detection [32], [60], interviewing [1], [4], [61], and risk assessment [49], [74], [76], [78], [91] under umbrella of justice and privacy issues [10], [15], [19]. In the International Air Transport Association (IATA) roadmap [36], four pillars of the checkpoint of the future are identified: computation intelligence, operations, infrastructure, and economic measures.…”

Biometric‐enabled watchlists technology

“…This conceptual projection includes the following dimensions: Inference engine : Since the data about the persons of interest that is stored in the watchlist is often incomplete, uncertain, imperfect, fragmentary, and conflicting [2, 3], a powerful inference engine is needed at various phases of gathering, evidence accumulation, control, fusion, recognition, tracking in the surveillance network, as well as risk assessment and prediction. These inferences can be executed by Bayesian networks [18] and their extensions such as dynamic Bayesian networks, as well as methods for dealing with conflicting information [19], and deep learning techniques based on neural networks [20]. Technological components, including artificial intelligence and computational intelligence tools, should be developed, implemented, and deployed under fundamental social constraints and regulations. Social embedding : The watchlist utilises mechanisms for embedding in social infrastructure such as exploring various databases, local security resources, big data analysis, and possibilities of surveillance networks for identification and tracking. Interview supporting technology : Motivated by the fact that the traveller's cooperation is a crucial factor for improving the watchlist performance, the watchlist screening should be integrated into interview supporting machines – a well identified trend for deception detection as applied to e‐border infrastructure. Countermeasures : The watchlist can be vulnerable to attacks on its integrity.…”

“…Inference engine : Since the data about the persons of interest that is stored in the watchlist is often incomplete, uncertain, imperfect, fragmentary, and conflicting [2, 3], a powerful inference engine is needed at various phases of gathering, evidence accumulation, control, fusion, recognition, tracking in the surveillance network, as well as risk assessment and prediction. These inferences can be executed by Bayesian networks [18] and their extensions such as dynamic Bayesian networks, as well as methods for dealing with conflicting information [19], and deep learning techniques based on neural networks [20]. Technological components, including artificial intelligence and computational intelligence tools, should be developed, implemented, and deployed under fundamental social constraints and regulations.…”